Measuring Multiprocessor System Activity

Benefits of Scaling

The process of adding processors to your system to achieve higher throughput is called scaling. Applications that benefit from multiprocessor configurations are typically those that are processor-intensive, such as database servers, Web servers, and active file and print servers. Processor-intensive applications that use multiple processes or are multithreaded with asynchronous execution are also well suited to multiprocessor systems. Systems requiring heavy computation capability, including detailed calculation for scientific or financial applications, complex graphic rendering, computer aided design (CAD)– based modeling, or electrical-engineering design might also demand multiprocessor systems.

To understand the degree of benefit you can attain from scaling to multiple processors, determine the scale factor. The scale factor is a measurement of the increase in throughput you can expect to achieve. You can determine the scale factor of the system by comparing the throughput of one processor to the throughput of multiple processors.

The formula for computing scale factor is:

Number of transactions per second on n processors ÷ Number of transactions per second on 1 processor

For example, if a single-processor system is using 100 percent of the processor handling 100 transactions per second and, with the addition of three processors, the resulting four-processor system can handle 320 transactions per second, the scale factor of the system is 3.2.